A Phaseless Source Reconstruction Method Based on Adam Optimization Algorithm Combined with Regularization.

Area of Science:

• Electromagnetics
• Computational Physics
• Signal Processing

Background:

• Traditional least squares methods for equivalent dipole solutions in electromagnetic inverse problems suffer from ill-conditioned matrices, leading to poor accuracy and noise sensitivity.
• Existing optimization algorithms often get trapped in local optima during iterative reconstruction processes.

Purpose of the Study:

• To develop a stable and accurate phaseless source reconstruction method for equivalent dipole problems.
• To overcome the limitations of traditional least squares and existing iterative optimization algorithms.

Main Methods:

• Proposed a novel method combining the Adam optimization algorithm with L2 regularization for phaseless source reconstruction.
• Adam optimization is used to circumvent direct inversion of ill-conditioned matrices, improving near-field source reconstruction accuracy.
• L2 regularization is incorporated to further suppress local optima and enhance the anti-noise performance of the dipole solution.

Main Results:

• The proposed method demonstrates stable and accurate solutions for equivalent dipole sources.
• Significantly improved accuracy in near-field source reconstruction compared to traditional methods.
• Enhanced robustness against noise and effective avoidance of local optima during iteration.

Conclusions:

• The combined Adam optimization and L2 regularization method provides a stable and effective solution for phaseless equivalent dipole reconstruction.
• The approach offers superior accuracy, anti-noise performance, and convergence properties for electromagnetic inverse problems.
• Simulation and experimental results validate the effectiveness of the proposed technique.